Routing and Wavelength Assignment (RWA) is one of the key issues in Intelligent Optical Network (ION). Dueto the Wavelength Continuity Constraint (WCC), RWA algorithms need to introduce the wavelength converter to solve the wavelength conflict. An optical network dynamic RWA algorithm based on improved ant colony optimization is proposed. The link idle ratio was introduced into the traditional ant transition probability as a constraint and a novel wavelength converter allocation strategy was discussed. Compared the blocking probability and resources utilization through the regular MESH and NSFNET network with numerical simulation, the result shows that in the case of the same traffic intensity, the blocking probability and channel utilization performance of the improved algorithm are improved significantly, compared with Dijkstra + First-Fit algorithm with wavelength convertors.